large model

In the early morning of December 7, Beijing time, Google broke the dawn and finally released the long-awaited AI model Gemini, a mysterious new model that Google has high hopes for, and is regarded as a "killer app" to deal with OpenAI's GPT-4. Living up to the expectations, Gemini has brought amazing results on its debut: it has been recognized as a "killer app" against OpenAI GPT-4 in the MMLU (Massive Multi...

Google recently announced the launch of Translatotron 3, a new AI model with an exciting feature: simultaneous translation. Unlike traditional speech-to-speech models, Translatotron 3 enables translation without the need for text conversion. This undoubtedly opens up new paths for cross-linguistic communication across the globe. ...

This past Tuesday Stability AI released the latest version of its Stable Video Diffusion model, a powerful technology capable of converting images into video. With this update, Stability AI also added support for the ComfyUI tool, a graphical user interface designed to use the graph/node interface...

Meta recently quietly released the Llama 2 Long AI model, a development that has garnered widespread attention in the AI field. The model demonstrated superior performance on a number of tasks, outperforming predecessors such as GPT-3.5 Turbo and Claude 2. Meta researchers boosted this by improving the training methodology and coding techniques...

Microsoft recently released a new tool called AutoGen, designed to help developers create complex applications based on the Large Language Model (LLM). The AutoGen tool has the following key features: Provides a multi-agent session framework as a high-level abstraction that enables developers to easily build LLM workflows. Provides...

Recently, the IDEA Institute's Seal of Approval team joined the open source trend of AI macromodeling with the release of its latest code macromodel, Ziya-Coding-34B-v1.0.This model achieved an excellent score of 75.5 on the HumanEval Pass@1 review, surpassing the GPT-4. Ziya-Coding-34B- ...

China Telecom led the innovation in the field of information and communication at the recent Network Big Model Technology Seminar Forum, releasing the first network big model - "Qiming". Xia Bing, deputy general manager of China Telecom, said that big model technology has become an important trend in the field of artificial intelligence, and domestic big models are showing diversified development. China ...

In the field of artificial intelligence, the cost of training large models has always been a challenge. However, a study by the Beijing Zhiyuan Artificial Intelligence Research Institute and the Institute of Computing Technology of the Chinese Academy of Sciences, among others, breaks this status quo. With a budget of just $100,000, they trained a brand new 1,000,000,000,000 gerunds with 101,000,000,000 gerunds...

Meta, which has been aggressively snapping up the AI training chip Nvidia H100 and setting up data centers, plans to develop a new AI system (i.e., a large-scale language model) in early 2024 that will rival OpenAI's GPT-4 in terms of capabilities, according to a report in the Wall Street Journal. The model is expected to be more powerful than Llam...

Chianxin Group recently released Q-GPT Security Robot and Big Model Defender, aiming to provide comprehensive quadruple protection for enterprises when using big models by means of security risk discovery, access control, and data leakage control, so as to reduce data security risks. In order to prevent employees from feeding sensitive data to large models lead...

Xiaomi's mobile voice assistant Xiao Ai Classmate will continue to open up large model internal testing slots next week to give more Mi users the chance to experience the upgraded Xiao Ai Classmate. Through the registration review, users will receive the Xiaomi community station message push to get the qualification to participate in the internal test. It is understood that the second batch of internal testing quota will cover including to...

Recently, Aliyun's Magic Hitch community announced the open source of a visual language model called Qwen-VL. The model is based on Qwen-7B, a 7 billion parameter model of Tongyi Qianqi, as a base language model, which has the ability of graphic input and multimodal information understanding, and is more suitable for the needs of Chinese users. Qwen-VL is based on Qwen-7B...

The domestic big model market has once again ushered in a new participant, this time a math-specific big model built specifically for the field of mathematics. According to the heart of the machine on August 24th, Tian Mi, CTO of the good future, announced in the 20th anniversary of the good future live event that the good future independent research and development of hundreds of billions of large models in the field of mathematics MathGPT is officially online...

Based on the self-developed giant language model, Racer's AI dialog function officially debuted on Android platform Recently, according to several media reports, Racer announced that its application based on the self-developed giant language model has made new progress, in which the "Racer's AI Dialog" function has been open for internal testing on the Android side of the application. The function has been open for internal testing in the Android app...